Calculating Sample Sizes in the Presence of Confounding Variables
本文针对研究设计中如何确定所需观测数量的问题，提出一种基于广义线性模型的通用理论，以在存在混杂变量时计算样本量，并通过三个医学研究实例展示应用。
A major issue in the design of many studies is to determine the “required” number of observations. Much of the statistical literature on sample size estimation, particularly for medical studies, is devoted to considering a very simple design, involving the testing of the difference between proportions in two groups. Accommodation of potential confounding variables at the design stage is more difficult. This paper outlines a general theory based on generalized linear models for accommodating such variables. The theory is applied to three examples from medical research.